Heuristic Optimization-Assisted Dilated Convolution Neural Network With Gated Recurrent Unit for Channel Estimation in NOMA-OFDM System

The "Non-Orthogonal Multiple Access (NOMA)" strategies has been recently identified as a successful way to increase spectrum effectiveness and reliability of the system. NOMA enables numerous users to share identical blocks of resources with varying power distribution factors. Thus, everyone can utilize equal resource structures, resulting in improved spectral effectiveness. "Orthogonal Frequency-Division Multiplexing (OFDM)" represents a well-known multi-carrier broadcast technique that provides a high data prevalence, excellent spectral effectiveness, and resilience against network band selection for present and future wideband wireless connections. However, channel estimation remains one of the most important challenges in OFDM based communication. Pilot-aided estimation of channels is less difficult and performs better than blind channel estimation; therefore, it is commonly employed in real-world systems. However, inadequate Successive Interference Cancellation (SIC) may have an impact on NOMA effectiveness. To help with channel estimates and signal identification in NOMA structures, deep learning approaches were developed. In this paper, we introduce a new approach called Adaptive Dilated Convolutional Neural Networks with Gated Recurrent Unit Layer (ADCNN-GRU). Initially, the required signals are collected from the standard data resource. The input signal is then subjected to the developed ADCNN-GRU approach, which is the combination of the Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU) with an additional dilated layer. Here the input signals are extracted and infer the signal at the receiver terminal. The loss functions in the model are optimized by using the Improved Pelican Optimization Algorithm (IPOA). The performance of the developed approach is determined by conducting the simulation experiment. The result showed that the developed approach outperformed than traditional models.